PERFORMANCE, EXPLAINED
How player ratings work
Ratings measure verified high-school game performance. Recruiting stars, offers, popularity, school reputation and team wins do not earn player points. Statistical leaderboards remain separate from performance rankings.
Different roles, different evidence
Each game combines production, efficiency, impact plays and a bounded opponent adjustment. QBs balance production and efficiency equally. Defensive impact includes sacks, tackles for loss, interceptions, defended passes and forced fumbles. Tackles alone cannot establish elite defensive performance. Missing metrics stay unknown.
QB model
production: 35% · efficiency: 35% · opponent: 20% · impact: 10%
production
- passYards: relative weight 0.85
- rushYards: relative weight 0.15
efficiency
- completionRate: relative weight 0.25
- passYPA: relative weight 0.3
- adjustedYPA: relative weight 0.45
impact
- passTD: relative weight 0.45
- rushTD: relative weight 0.15
- passINT: relative weight 0.4 · lower is better
Prior strength: 1.25 equivalent games. Required inputs: passYards, passAttempts.
RB model
production: 40% · efficiency: 30% · opponent: 20% · impact: 10%
production
- rushYards: relative weight 0.7
- recYards: relative weight 0.2
- receptions: relative weight 0.1
efficiency
- rushYPC: relative weight 0.7
- yardsPerTouch: relative weight 0.3
impact
- rushTD: relative weight 0.65
- recTD: relative weight 0.15
- fumblesLost: relative weight 0.2 · lower is better
Prior strength: 1.25 equivalent games. Required inputs: rushYards, carries.
WR model
production: 40% · efficiency: 30% · opponent: 15% · impact: 15%
production
- recYards: relative weight 0.7
- receptions: relative weight 0.3
efficiency
- yardsPerReception: relative weight 0.55
- yardsPerTarget: relative weight 0.25
- catchRate: relative weight 0.2
impact
- recTD: relative weight 0.85
- fumblesLost: relative weight 0.15 · lower is better
Prior strength: 1.5 equivalent games. Required inputs: recYards, receptions.
TE model
production: 40% · efficiency: 25% · opponent: 15% · impact: 20%
production
- recYards: relative weight 0.65
- receptions: relative weight 0.35
efficiency
- yardsPerReception: relative weight 0.6
- yardsPerTarget: relative weight 0.4
impact
- recTD: relative weight 0.85
- fumblesLost: relative weight 0.15 · lower is better
Prior strength: 1.5 equivalent games. Required inputs: recYards, receptions.
DL model
production: 30% · efficiency: 10% · opponent: 15% · impact: 45%
production
- tackles: relative weight 0.5
- tfl: relative weight 0.5
efficiency
- pressureRate: relative weight 1
impact
- sacks: relative weight 0.5
- forcedFumbles: relative weight 0.25
- blockedKicks: relative weight 0.25
Prior strength: 1.5 equivalent games. Required inputs: tackles, sacks.
EDGE model
production: 25% · efficiency: 15% · opponent: 15% · impact: 45%
production
- tackles: relative weight 0.3
- tfl: relative weight 0.6
- pressures: relative weight 0.1
efficiency
- pressureRate: relative weight 1
impact
- sacks: relative weight 0.65
- forcedFumbles: relative weight 0.25
- blockedKicks: relative weight 0.1
Prior strength: 1.5 equivalent games. Required inputs: tackles, sacks.
LB model
production: 30% · efficiency: 10% · opponent: 15% · impact: 45%
production
- tackles: relative weight 0.5
- soloTackles: relative weight 0.25
- tfl: relative weight 0.25
efficiency
- coverageStopRate: relative weight 1
impact
- sacks: relative weight 0.25
- defINT: relative weight 0.2
- pbu: relative weight 0.15
- forcedFumbles: relative weight 0.25
- recoveries: relative weight 0.1
- tfl: relative weight 0.2
- defTD: relative weight 0.1
Prior strength: 1.5 equivalent games. Required inputs: tackles.
CB model
production: 15% · efficiency: 40% · opponent: 15% · impact: 30%
production
- pbu: relative weight 0.65
- tackles: relative weight 0.15
- coverageStops: relative weight 0.2
efficiency
- coverageStopRate: relative weight 0.65
- yardsAllowedPerTarget: relative weight 0.35 · lower is better
impact
- pbu: relative weight 0.25
- defINT: relative weight 0.3
- forcedFumbles: relative weight 0.2
- defTD: relative weight 0.25
Prior strength: 2 equivalent games. Required inputs: pbu.
S model
production: 30% · efficiency: 20% · opponent: 15% · impact: 35%
production
- tackles: relative weight 0.55
- pbu: relative weight 0.3
- tfl: relative weight 0.15
efficiency
- coverageStopRate: relative weight 1
impact
- defINT: relative weight 0.35
- forcedFumbles: relative weight 0.3
- defTD: relative weight 0.2
- tfl: relative weight 0.15
Prior strength: 1.5 equivalent games. Required inputs: tackles.
K model
production: 15% · efficiency: 50% · opponent: 5% · impact: 30%
production
- fgMade: relative weight 0.65
- xpMade: relative weight 0.35
efficiency
- fgRate: relative weight 0.65
- xpRate: relative weight 0.35
impact
- fgLong: relative weight 0.7
- touchbacks: relative weight 0.3
Prior strength: 2 equivalent games. Required inputs: fgAttempts, fgMade.
P model
production: 30% · efficiency: 40% · opponent: 5% · impact: 25%
production
- puntAverage: relative weight 0.7
- puntNet: relative weight 0.3
efficiency
- inside20Rate: relative weight 0.65
- puntTouchbackRate: relative weight 0.35 · lower is better
impact
- puntLong: relative weight 0.4
- inside20: relative weight 0.6
Prior strength: 2 equivalent games. Required inputs: punts, puntYards.
Returner model
production: 25% · efficiency: 40% · opponent: 10% · impact: 25%
production
- returnYards: relative weight 1
efficiency
- returnAverage: relative weight 1
impact
- returnTD: relative weight 1
Prior strength: 2 equivalent games. Required inputs: returnAttempts, returnYards.
Quiet cornerback box scores cannot establish poor coverage. Without target data, low counts are neutral evidence and confidence remains Low. Kicking, punting and return efficiency need minimum attempts. Offensive linemen show “Insufficient standardized performance data.” ATH or generic DB listings need a verified specific position.
Competition and repeatability
Opponent strength uses the same internal team power for NJ and out-of-state opponents. It contributes at most six internal game-score points in either direction. Public stars never enter the calculation. Overall team strength is used until reliable offense/defense splits are available.
The season score is 80% weighted full-season performance, 15% recent three-game performance and 5% capped peak. Game weights step down modestly: 1.00, 0.96, 0.92. The peak cannot exceed the full-season mean by more than 12 internal points.
A stable 0–100 scale
Reported metrics are compared within position using robust median and spread references. Extreme metric scores are clipped, and implausible reports are held for review. The season score is standardized within position, then stabilized with a position-average prior of 1.25–2 equivalent games. Only the number of usable games determines this prior weight; missing-field percentages do not increase it.
The stabilized score is compared with a frozen position reference, then mapped through a smooth, fixed public rating curve. References come from the current eligible covered population where at least 12 players exist. Sparse positions use fixed role baselines. The reference persists for this season and model; a new week or new leaderboard maximum does not rescale existing performances. Historical validation remains limited.
| Position percentile | Public rating |
|---|---|
| 10th | 47 |
| 30th | 59 |
| 50th | 69 |
| 70th | 79 |
| 80th | 84 |
| 90th | 89 |
| 95th | 93 |
| 97th | 95 |
| 99th | 97.5 |
Rating and confidence are separate
Coverage and source quality describe how much we know. They never multiply the rating. Missing games are not zero performances. Reports are marked Complete, Partial, Summary only or Missing. A dominant one-game sample may have a high rating, Low confidence and no official rank.
Position rank requires two verified usable appearances, 50% game coverage, 35% position-input coverage and known opponent powers. Overall requires three usable games and 75% game coverage. Box-score-only cornerback ratings do not qualify for Overall. Two-game position ratings remain labeled Provisional.
Confidence combines games, reporting coverage and available fields. Its uncertainty index is descriptive, not a validated probability. Sparse positions and schools with missing reports need particular care. These lists represent covered players, not every varsity player in New Jersey.
Two-way players and Overall
Each supported role receives a separate rating. Overall compares position-normalized performance, using the strongest qualified role plus 30% of a secondary role’s above-average standardized contribution, capped at 0.6 standardized points. Ratings are never added together. A player appears once in Overall.
Sources, corrections and history
Game reports must match completed, undisputed games and the correct team. Approved evidence takes priority; weaker, older or conflicting imports cannot overwrite stronger verified stats. Every game retains its source, checked time, source confidence and missing fields. Stat corrections trigger a new calculation through game ratings, season scores, percentiles and all derived ranks.
Published season totals are never added to game logs or used as substitutes for game-level ratings. Player of the Week uses position-normalized game performance. History preserves real weekly snapshots; movement begins once an earlier week exists for this model. Past ranks are not invented.
Current opponent powers may differ from pregame powers. Advanced coverage, pressures, targets, field-goal distances and blocking remain incomplete. Historical 2025 opponent powers must be supplied before 2025 performance rankings qualify. Automatic source updates remain paused.
Model version: performance-v2.0.1.